[FEA] set shuffle numThreads dynamically by default

[revans2]

Is your feature request related to a problem? Please describe.
The auto tuner code currently sets the number of shuffle threads to 28 for CSPs and 1.5 * numExecutorCores for on prem environments. We should look into which if either of these we might want to use in the plugin by default, or if there is something better we can use to figure this out. We do know the number of executor cores, because if the parallelism is less than the number of executor cores, then a warning is output. We should run some performance tests to verify that whatever change we make is better generally.